diff options
author | patacongo <patacongo@42af7a65-404d-4744-a932-0658087f49c3> | 2012-02-21 22:00:28 +0000 |
---|---|---|
committer | patacongo <patacongo@42af7a65-404d-4744-a932-0658087f49c3> | 2012-02-21 22:00:28 +0000 |
commit | 5c775298fca577723a497605f0f4dec4df9fa05a (patch) | |
tree | 42953b95266cf4011d01d1580f5c0eff56518337 /nuttx/configs/stm3240g-eval/README.txt | |
parent | c56a18fb38e4d9d11b6799eef395a6dc667750e7 (diff) | |
download | px4-nuttx-5c775298fca577723a497605f0f4dec4df9fa05a.tar.gz px4-nuttx-5c775298fca577723a497605f0f4dec4df9fa05a.tar.bz2 px4-nuttx-5c775298fca577723a497605f0f4dec4df9fa05a.zip |
STM32 SDIO DMA: Ignore DMA FIFO errors; these seem to be bogus. SD multiple block transfers result in CRC errors; avoid them by using smaller FTP buffer sizes
git-svn-id: svn://svn.code.sf.net/p/nuttx/code/trunk@4410 42af7a65-404d-4744-a932-0658087f49c3
Diffstat (limited to 'nuttx/configs/stm3240g-eval/README.txt')
-rwxr-xr-x | nuttx/configs/stm3240g-eval/README.txt | 22 |
1 files changed, 22 insertions, 0 deletions
diff --git a/nuttx/configs/stm3240g-eval/README.txt b/nuttx/configs/stm3240g-eval/README.txt index e4902947e..3c4f7e77c 100755 --- a/nuttx/configs/stm3240g-eval/README.txt +++ b/nuttx/configs/stm3240g-eval/README.txt @@ -737,6 +737,28 @@ Where <subdir> is one of the following: Then DMA works fine. The downside is, of course, is that we lose 64Kb of precious SRAM. + 5. Another SDIO/DMA issue. This one is probably a software bug. This is + the bug as stated in the TODO list: + + "If you use a large I/O buffer to access the file system, then the + MMCSD driver will perform multiple block SD transfers. With DMA + ON, this seems to result in CRC errors detected by the hardware + during the transfer. Workaround: Use I/O buffers less the 1024 + bytes." + + For this reason, CONFIG_FTPD_DATABUFFERSIZE=512 appears in the defconfig + file. + + 6. Another DMA-related concern. I see this statement in the reference + manual: "The burst configuration has to be selected in order to respect + the AHB protocol, where bursts must not cross the 1 KB address boundary + because the minimum address space that can be allocated to a single slave + is 1 KB. This means that the 1 KB address boundary should not be crossed + by a burst block transfer, otherwise an AHB error would be generated, + that is not reported by the DMA registers." + + There is nothing in the DMA driver to prevent this now. + ostest: ------ This configuration directory, performs a simple OS test using |